- Email: [email protected]
Analysis of heavy rainfall events in North Rhine–Westphalia with radar and raingauge data
Atmospheric Research 77 (2005) 337 – 346 www.elsevier.com/locate/atmos
Analysis of heavy rainfall events in North Rhine–Westphalia with radar and raingauge data Markus Jessena,T, Thomas Einfalta, Andre´ Stoffera, Bernd Mehligb a
einfalt&hydrotec GbR, Wakenitzmauer 33, D-23552 Luebeck, Germany b Landesumweltamt NRW, Wallneyer Str. 6, D-45133 Essen, Germany
Received 16 March 2004; received in revised form 4 November 2004; accepted 5 November 2004
Abstract Five heavy rainfall events were investigated with radar and raingauge data. Special attention was paid to quality check and adjustment of radar data. Attenuation effects could be observed on both, CBand and on X-Band radar. Adjustment of radar data to raingauge values turned out to be difficult in the vicinity of heavy local rain cells. Four adjustment methods were analysed and radar data from different radar stations were compared. As a further result of this project, the spatial extent of the precipitation fields was identified by adjusted radar data and compared to raingauge data. For each rainfall event, radar derived accumulated rainfall images and catchment time series were produced. D 2005 Elsevier B.V. All rights reserved. Keywords: Radar; Rainfall; Raingauge; Damage; Quality control; Adjustment; Attenuation
1. Introduction In the summers of 2001 and 2002 heavy, rainfall events caused large damages in five locations of North Rhine–Westphalia (Fig. 1). These events were very localised and caused severe damage such as landslides and flooding.
T Corresponding author. Tel.: +49 451 7027333; fax: +49 451 7027339. E-mail addresses: [email protected] (M. Jessen), [email protected] (B. Mehlig). 0169-8095/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2004.11.031
338
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
Fig. 1. Affected locations and used radar stations.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
339
The North Rhine–Westphalia State Environment Agency requested the analysis of the five events with the following objectives: ! ! ! ! !
quantification of the precipitation with raingauge and radar data, determination of the spatial extent of the extreme precipitation, comparison between the different radar locations and measurement methods, analysis of the damage locations in relation to the reasons of damage, preparation of time series for areal precipitation.
2. Available data The following data were used for the analysis of the five rainfall events: ! ! ! ! !
data from one up to four radars, data of up to 91 raingauge stations with daily and continuous values, flow gauge measurement data, geographic data (e.g. catchment boundaries, stream network, digital terrain model), reports from the German Weather Service (DWD) and other institutions.
2.1. Weather radar The German Weather Service (DWD) is the operator of the three C-band radar stations Flechtdorf, Essen and Neuheilenbach used in this project. The fourth radar station in the project (Bonn, X-Band radar) is operated by the Meteorological Institute of the University of Bonn. Data from both sources were 5-min data with 256 intensity classes with 0.5-dBZ resolution. Radar data from the DWD were based on the DX-product (DWD, 1997), a PPI data set being characterised by: ! 1 km 18 (for further processing converted to 1 km2 spatial resolution), ! 128 km range, ! 0.88 beam elevation. The X-band weather radar in Bonn has the following characteristics: ! 250 m spatial resolution, aggregated to 500 500 m pixels, ! 50 km range (2.58 elevation angle) and 100 km range (1.58 elevation angle). For some of the events, data from up to all four radar locations were available. 2.2. Raingauges The density of available raingauges varied for the five events and locations. For the event with the highest density of raingauges, data from 91 continuous and daily stations could be used. In one case the analysis was only based on 27 stations.
340
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
3. Quality check An intensive quality check of the radar and raingauge data (Einfalt et al., 2004a,b) made obvious the following points with influence on the final results: ! ! ! !
the correct justification of the radar images to the north, the heights of the radar measurement and of the location of the raingauges, the accuracy of the raingauge coordinates, the identification of attenuation as a result of a comparison between different radar locations.
The issues of sampling errors of the radar (i.e. the representativeness of a radar grid measurement, see Einfalt et al., 2004a,b) and of the sampling problem when comparing raingauge and radar measurements (see Joss and Germann, 2004) have not been investigated in this study. 3.1. North justification Each radar is justified to the geographic north. The geographic data (e.g. catchment boundaries, etc.) in Germany are based on the Gauss–Krueger rectangular projection where the north justification is represented by the middle meridian of the Gauss–Krueger zone. If the radar station is not positioned on the middle meridian, there is a difference in the justification of up to 18. This problem increases with growing distance from the radar and may produce a difference in location of several pixels at the far range. For this project, the north justification of the radar images was adapted to the north justification of the geographic data. 3.2. Measurement elevation and range The height of the radar measurement is determined by the distance between radar station and measurement point, the geographical height of the radar station, the elevation angle and the beamwidth of the radar beam. Depending on these attributes, measurements for one location from different radar stations at different heights above ground are obtained. Therefore, different measurement values for the same location are possible if two or more radar stations and the raingauge at the ground are compared. An important source for these differences is the variability of the rainfall intensity in different heights (see, e.g. Meischner, 2003 for more details). 3.3. Accuracy of raingauge coordinates Coordinates of raingauge locations can be inexact if the raingauge station is moved or the location is not exactly known. The size of a radar pixel is 1 kM (DWD weather radar) or 250 M for the radar in Bonn. Therefore a little inaccuracy in the coordinates of the raingauges can cause a significant problem when comparing radar and raingauge data.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
341
Fig. 2. Time series with mixed effects of attenuation and sampling errors (event of 3 May 2001).
3.4. Attenuation High precipitation intensities attenuate the radar beam, so that the value of the reflected beam behind the high precipitation area appears to be lower than it is in reality. Attenuation is a problem with C-band and especially X-band radars and can be detected in single images. In this project the effects of attenuation have been observed in different regions of the available radar images. The effects can also be seen in the corresponding time series of radar data at the Marienfeld gauge location in Fig. 2 (data bulk adjusted with the Marienfeld values). While both radars from the DWD, Flechtdorf (black, dashed) and Essen (gray, dashed) are attenuated over station Marienfeld (location in Fig. 1) at 16.15 and 14.45 h, respectively, the measurements from the Bonn radar (gray) are in good agreement with the ones of the raingauge (black). The attenuation effect had been analysed in details by PPI data of the concerned radars (not presented in this paper). The quantification of the attenuation effect is difficult. One reason for this is that attenuation is associated with heavy precipitation, which at the same time is responsible for high spatial variation in rainfall, leading to sampling problems (inhomogeneous beam filling, questions of representativeness of a raingauge in space, etc.). So the part attributed to the attenuation effect in the data as compared to the other effects cannot be straightforward determined.
4. Adjustment of radar data Radar measurements are rainfall measurements with an indirect method which is not linearly related to rainfall rate. Therefore in practice point measurements are used for the best possible determination of the precipitation volume. There are many possible adjustment and correction techniques to obtain adjusted radar data (for detailed references,
342
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
see Meischner, 2003). Here, uncorrected radar data with an adjustment factor matrix were used as standard method. The following adjustment methods were analysed: ! event-wise constant adjustment factor-constant in space and time, ! adjustment factor matrix (Wilson and Brandes, 1979)—variable in space, constant in time over the event, ! adjustment factor matrix based on the neighbourhood of nine pixels, ! adjustment factor matrix after using a correction method for attenuation. Criteria for the evaluation of the methods were: ! the comparison of the shapes of produced pixel time series to the shape of the corresponding raingauge time series, ! the comparison of radar derived rainfall amounts to raingauge amounts (for the first method only), ! the relative homogeneity of the correction factor field, i.e. there should be no obvious outlayers. 4.1. Event-wise constant adjustment factor The quotient of the accumulated measured precipitation at the raingauge stations and the accumulated radar data at the same locations produces the factor. The time period for the sum is based on the period of the interesting event. The results of this method were not acceptable because of ! a high spatial variability in the relationship between rainfall intensity and radar reflectivity due to the different drop spectra of the mainly small cells and their surrounding rainfall areas, ! the influence of radar beam attenuation. 4.2. Adjustment factor matrix For this method the individual adjustment factors are determined at the locations of the raingauge stations. Between the raingauge locations, an interpolation with the adjustment factors from the three nearest stations based on the quadratic reciprocal distance was used. As a minimum requirement, the measured precipitation from raingauge and radar had to be more than 1 mm. The results were plausible. Problems could be observed when raingauge locations were at the pixel boundary and radar data had a steep gradient at this location. 4.3. Adjustment factor matrix based on the neighbourhood of nine pixels The method with an adjustment factor matrix is problematic if the precipitation is very variable and the raingauge station is located near the boundary of two pixels. Fig. 3 (radar data from DWD) is showing such an example, numbers are giving unadjusted radar measurements in mm, the raingauge Hennef–Stadt Blankenberg had
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
343
Fig. 3. Spatial variability of precipitation.
collected 73.5 mm. In this case it may be better to calculate the adjustment factors from the mean value of the pixel at the raingauge location and the eight pixels around this location. The results were mainly plausible. Problems occurred with very small scale extreme rainfall which was averaged over the nine pixel area and sometimes strongly underestimated extreme rainfall observed at the raingauge. This, in turn, led to high correction factors which were much higher than other ones in the close vicinity. 4.4. Adjustment factor matrix after using a correction method for attenuation Because of the mentioned effects of attenuation, a combined detection and correction method was applied before the computation of the adjustment factor matrix. To improve the attenuated area behind high dBZ values (here identified by values z 48 dBZ), a cumulative gate-by-gate algorithm (Harrison et al., 2000) was used. By this method an iterative correction along each radar beam is performed. The algorithm has to be capped (in this example at the factor of two for the increase of the rainrate), because otherwise it tends to instability in the case of severe attenuation. Due to lack of mountains, other algorithms, e.g. making use of regular ground echoes returns (Sempere-Torres et al., 2001) were not applicable here. The results were not always plausible. One problem consisted in the large mean field bias between raingauge and radar measurements. This bias was significantly different for
344
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
the different radar stations. This leads to two effects: the comparability of the method for different radars is not given, and the iterative correction methods require a data set without bias between radar and raingauges. For the reasons, and due to the inherently unstable character of the correction procedure, the use of this method was not chosen for further work.
5. Results of the project In spite of the discussed shortcomings, the second method (adjustment factor method) was on average the best method (based on radar-raingauge comparison) and selected for the production of further results. The results show that the spatial extent of heavy precipitation areas (larger than 100 mm) can only be identified with radar data or an adequate quantity of raingauges (here present for one event, shown in Fig. 5). In three events the centre of the heavy precipitation area had only a spatial extent of less than 15 km2. For an overview of the spatial extent of heavy precipitation, images of accumulated rainfall were produced for each event. An example for the event in the region of Gummersbach is shown in Fig. 4 (adjusted radar data from DWD) and Fig. 5 (interpolated raingauge data). Not in all affected areas the raingauges could observe the maximum of precipitation
Fig. 4. Accumulated and adjusted radar image for the rainfall event of 3 May 2001.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
345
Fig. 5. Interpolated raingauge image for the rainfall event of 3 May 2001.
because of the low raingauge density. For each area the differences between the maximum value of adjusted radar data and the maximum raingauge value are presented in Table 1. An important influence on the occurrence of attenuation played the wavelength of the radar. As expected, the X-band radar had the largest problems with this effect. A quantification, how much of the observed reduction of rainfall as seen by single radars can be attributed to attenuation and how much to sampling errors or other effects, is difficult and has not been attempted.
Table 1 Maximum values of adjusted radar data and raingauge stations Raingauge station
Raingauge value [mm]
Maximum radar value [mm]
Area
Warburg–Welda Bad Driburg–Herste (daily value) Wiehl (daily value) Hennef–Stadt Blankenberg (daily value) Niederdielfen KLG Beckum KA (daily value) Soest KLG
74.5 81.8 110.5 73.5 115.4 86.0 35.1
100 110 138 104 134 96 64
Warburg Bad Driburg Gummersbach Eitorf Siegen Ahlen–Beckum Soest-Welver
346
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
Further results of this work are: ! preparation of catchment and pixel time series (time series for each radar pixel of the catchments), ! classification of the events on the basis of extreme value statistics (DWD, 1997), ! maps with the damage locations for each event, ! outlets, bridges and other buildings near rivers and creeks are the problematic points for damages.
6. Consequences Consequences for the use of radar data are: ! the quality of the radar images varies between the different radar stations and has to be checked and–where feasible–corrected carefully, ! for the adjustment of radar data, a high density of raingauge stations (continuous and daily values) is helpful. The produced catchment and pixel time series will be used in hydrological models for planning purposes and for scenario simulations of extreme rainfall.
References DWD, 1997. AKORD—Anwenderorientierte Organisation von Radardaten. Deutscher Wetterdienst, Offenbach. Einfalt, T., Arnbjerg-Nielsen, K., Faure, D., Jensen, N.E., Quirmbach, M., Vaes, G., Vieux, B., Golz, C., 2004. Towards a roadmap for use of radar rainfall data in urban drainage. Journal of Hydrology 299, 3 – 4. Einfalt, T., Golz, C., Jessen, M., 2004. Searching for rainfall truth: multisensor thunderstorm analysis. 3rd ERAD, Visby, Sweden, 6–10 September. Harrison, D.L., Driscoll, S.J., Kitchen, M., 2000. Improving precipitation estimates from weather radar using quality control and correction techniques. Meteorological Applications 6, 135 – 144. Joss, J., Germann, U., 2004. Hourly Assessment Factors between Radar and Raingauge, II. VOLTAIRE workshop, Ljubljana, October. Meischner, P., 2003. Weather Radar: Principles and Advanced Applications. Springer Verlag. Sempere-Torres, D., Sa´nchez-Diezma, R., Cordoba, M.A., Pascual, R., Zawadski, I., 2001. An operational methodology to control radar measurements stability from mountain returns. 30th Conf. On Radar Met., Munich, Germany, pp. 264 – 266. Wilson, J.W., Brandes, E.A., 1979. Radar Measurement of Rainfall—A Summary, vol. 60. American Meteorological Society, pp. 1048 – 1058.
Analysis of heavy rainfall events in North Rhine–Westphalia with radar and raingauge data Markus Jessena,T, Thomas Einfalta, Andre´ Stoffera, Bernd Mehligb a
einfalt&hydrotec GbR, Wakenitzmauer 33, D-23552 Luebeck, Germany b Landesumweltamt NRW, Wallneyer Str. 6, D-45133 Essen, Germany
Received 16 March 2004; received in revised form 4 November 2004; accepted 5 November 2004
Abstract Five heavy rainfall events were investigated with radar and raingauge data. Special attention was paid to quality check and adjustment of radar data. Attenuation effects could be observed on both, CBand and on X-Band radar. Adjustment of radar data to raingauge values turned out to be difficult in the vicinity of heavy local rain cells. Four adjustment methods were analysed and radar data from different radar stations were compared. As a further result of this project, the spatial extent of the precipitation fields was identified by adjusted radar data and compared to raingauge data. For each rainfall event, radar derived accumulated rainfall images and catchment time series were produced. D 2005 Elsevier B.V. All rights reserved. Keywords: Radar; Rainfall; Raingauge; Damage; Quality control; Adjustment; Attenuation
1. Introduction In the summers of 2001 and 2002 heavy, rainfall events caused large damages in five locations of North Rhine–Westphalia (Fig. 1). These events were very localised and caused severe damage such as landslides and flooding.
T Corresponding author. Tel.: +49 451 7027333; fax: +49 451 7027339. E-mail addresses: [email protected] (M. Jessen), [email protected] (B. Mehlig). 0169-8095/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2004.11.031
338
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
Fig. 1. Affected locations and used radar stations.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
339
The North Rhine–Westphalia State Environment Agency requested the analysis of the five events with the following objectives: ! ! ! ! !
quantification of the precipitation with raingauge and radar data, determination of the spatial extent of the extreme precipitation, comparison between the different radar locations and measurement methods, analysis of the damage locations in relation to the reasons of damage, preparation of time series for areal precipitation.
2. Available data The following data were used for the analysis of the five rainfall events: ! ! ! ! !
data from one up to four radars, data of up to 91 raingauge stations with daily and continuous values, flow gauge measurement data, geographic data (e.g. catchment boundaries, stream network, digital terrain model), reports from the German Weather Service (DWD) and other institutions.
2.1. Weather radar The German Weather Service (DWD) is the operator of the three C-band radar stations Flechtdorf, Essen and Neuheilenbach used in this project. The fourth radar station in the project (Bonn, X-Band radar) is operated by the Meteorological Institute of the University of Bonn. Data from both sources were 5-min data with 256 intensity classes with 0.5-dBZ resolution. Radar data from the DWD were based on the DX-product (DWD, 1997), a PPI data set being characterised by: ! 1 km 18 (for further processing converted to 1 km2 spatial resolution), ! 128 km range, ! 0.88 beam elevation. The X-band weather radar in Bonn has the following characteristics: ! 250 m spatial resolution, aggregated to 500 500 m pixels, ! 50 km range (2.58 elevation angle) and 100 km range (1.58 elevation angle). For some of the events, data from up to all four radar locations were available. 2.2. Raingauges The density of available raingauges varied for the five events and locations. For the event with the highest density of raingauges, data from 91 continuous and daily stations could be used. In one case the analysis was only based on 27 stations.
340
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
3. Quality check An intensive quality check of the radar and raingauge data (Einfalt et al., 2004a,b) made obvious the following points with influence on the final results: ! ! ! !
the correct justification of the radar images to the north, the heights of the radar measurement and of the location of the raingauges, the accuracy of the raingauge coordinates, the identification of attenuation as a result of a comparison between different radar locations.
The issues of sampling errors of the radar (i.e. the representativeness of a radar grid measurement, see Einfalt et al., 2004a,b) and of the sampling problem when comparing raingauge and radar measurements (see Joss and Germann, 2004) have not been investigated in this study. 3.1. North justification Each radar is justified to the geographic north. The geographic data (e.g. catchment boundaries, etc.) in Germany are based on the Gauss–Krueger rectangular projection where the north justification is represented by the middle meridian of the Gauss–Krueger zone. If the radar station is not positioned on the middle meridian, there is a difference in the justification of up to 18. This problem increases with growing distance from the radar and may produce a difference in location of several pixels at the far range. For this project, the north justification of the radar images was adapted to the north justification of the geographic data. 3.2. Measurement elevation and range The height of the radar measurement is determined by the distance between radar station and measurement point, the geographical height of the radar station, the elevation angle and the beamwidth of the radar beam. Depending on these attributes, measurements for one location from different radar stations at different heights above ground are obtained. Therefore, different measurement values for the same location are possible if two or more radar stations and the raingauge at the ground are compared. An important source for these differences is the variability of the rainfall intensity in different heights (see, e.g. Meischner, 2003 for more details). 3.3. Accuracy of raingauge coordinates Coordinates of raingauge locations can be inexact if the raingauge station is moved or the location is not exactly known. The size of a radar pixel is 1 kM (DWD weather radar) or 250 M for the radar in Bonn. Therefore a little inaccuracy in the coordinates of the raingauges can cause a significant problem when comparing radar and raingauge data.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
341
Fig. 2. Time series with mixed effects of attenuation and sampling errors (event of 3 May 2001).
3.4. Attenuation High precipitation intensities attenuate the radar beam, so that the value of the reflected beam behind the high precipitation area appears to be lower than it is in reality. Attenuation is a problem with C-band and especially X-band radars and can be detected in single images. In this project the effects of attenuation have been observed in different regions of the available radar images. The effects can also be seen in the corresponding time series of radar data at the Marienfeld gauge location in Fig. 2 (data bulk adjusted with the Marienfeld values). While both radars from the DWD, Flechtdorf (black, dashed) and Essen (gray, dashed) are attenuated over station Marienfeld (location in Fig. 1) at 16.15 and 14.45 h, respectively, the measurements from the Bonn radar (gray) are in good agreement with the ones of the raingauge (black). The attenuation effect had been analysed in details by PPI data of the concerned radars (not presented in this paper). The quantification of the attenuation effect is difficult. One reason for this is that attenuation is associated with heavy precipitation, which at the same time is responsible for high spatial variation in rainfall, leading to sampling problems (inhomogeneous beam filling, questions of representativeness of a raingauge in space, etc.). So the part attributed to the attenuation effect in the data as compared to the other effects cannot be straightforward determined.
4. Adjustment of radar data Radar measurements are rainfall measurements with an indirect method which is not linearly related to rainfall rate. Therefore in practice point measurements are used for the best possible determination of the precipitation volume. There are many possible adjustment and correction techniques to obtain adjusted radar data (for detailed references,
342
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
see Meischner, 2003). Here, uncorrected radar data with an adjustment factor matrix were used as standard method. The following adjustment methods were analysed: ! event-wise constant adjustment factor-constant in space and time, ! adjustment factor matrix (Wilson and Brandes, 1979)—variable in space, constant in time over the event, ! adjustment factor matrix based on the neighbourhood of nine pixels, ! adjustment factor matrix after using a correction method for attenuation. Criteria for the evaluation of the methods were: ! the comparison of the shapes of produced pixel time series to the shape of the corresponding raingauge time series, ! the comparison of radar derived rainfall amounts to raingauge amounts (for the first method only), ! the relative homogeneity of the correction factor field, i.e. there should be no obvious outlayers. 4.1. Event-wise constant adjustment factor The quotient of the accumulated measured precipitation at the raingauge stations and the accumulated radar data at the same locations produces the factor. The time period for the sum is based on the period of the interesting event. The results of this method were not acceptable because of ! a high spatial variability in the relationship between rainfall intensity and radar reflectivity due to the different drop spectra of the mainly small cells and their surrounding rainfall areas, ! the influence of radar beam attenuation. 4.2. Adjustment factor matrix For this method the individual adjustment factors are determined at the locations of the raingauge stations. Between the raingauge locations, an interpolation with the adjustment factors from the three nearest stations based on the quadratic reciprocal distance was used. As a minimum requirement, the measured precipitation from raingauge and radar had to be more than 1 mm. The results were plausible. Problems could be observed when raingauge locations were at the pixel boundary and radar data had a steep gradient at this location. 4.3. Adjustment factor matrix based on the neighbourhood of nine pixels The method with an adjustment factor matrix is problematic if the precipitation is very variable and the raingauge station is located near the boundary of two pixels. Fig. 3 (radar data from DWD) is showing such an example, numbers are giving unadjusted radar measurements in mm, the raingauge Hennef–Stadt Blankenberg had
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
343
Fig. 3. Spatial variability of precipitation.
collected 73.5 mm. In this case it may be better to calculate the adjustment factors from the mean value of the pixel at the raingauge location and the eight pixels around this location. The results were mainly plausible. Problems occurred with very small scale extreme rainfall which was averaged over the nine pixel area and sometimes strongly underestimated extreme rainfall observed at the raingauge. This, in turn, led to high correction factors which were much higher than other ones in the close vicinity. 4.4. Adjustment factor matrix after using a correction method for attenuation Because of the mentioned effects of attenuation, a combined detection and correction method was applied before the computation of the adjustment factor matrix. To improve the attenuated area behind high dBZ values (here identified by values z 48 dBZ), a cumulative gate-by-gate algorithm (Harrison et al., 2000) was used. By this method an iterative correction along each radar beam is performed. The algorithm has to be capped (in this example at the factor of two for the increase of the rainrate), because otherwise it tends to instability in the case of severe attenuation. Due to lack of mountains, other algorithms, e.g. making use of regular ground echoes returns (Sempere-Torres et al., 2001) were not applicable here. The results were not always plausible. One problem consisted in the large mean field bias between raingauge and radar measurements. This bias was significantly different for
344
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
the different radar stations. This leads to two effects: the comparability of the method for different radars is not given, and the iterative correction methods require a data set without bias between radar and raingauges. For the reasons, and due to the inherently unstable character of the correction procedure, the use of this method was not chosen for further work.
5. Results of the project In spite of the discussed shortcomings, the second method (adjustment factor method) was on average the best method (based on radar-raingauge comparison) and selected for the production of further results. The results show that the spatial extent of heavy precipitation areas (larger than 100 mm) can only be identified with radar data or an adequate quantity of raingauges (here present for one event, shown in Fig. 5). In three events the centre of the heavy precipitation area had only a spatial extent of less than 15 km2. For an overview of the spatial extent of heavy precipitation, images of accumulated rainfall were produced for each event. An example for the event in the region of Gummersbach is shown in Fig. 4 (adjusted radar data from DWD) and Fig. 5 (interpolated raingauge data). Not in all affected areas the raingauges could observe the maximum of precipitation
Fig. 4. Accumulated and adjusted radar image for the rainfall event of 3 May 2001.
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
345
Fig. 5. Interpolated raingauge image for the rainfall event of 3 May 2001.
because of the low raingauge density. For each area the differences between the maximum value of adjusted radar data and the maximum raingauge value are presented in Table 1. An important influence on the occurrence of attenuation played the wavelength of the radar. As expected, the X-band radar had the largest problems with this effect. A quantification, how much of the observed reduction of rainfall as seen by single radars can be attributed to attenuation and how much to sampling errors or other effects, is difficult and has not been attempted.
Table 1 Maximum values of adjusted radar data and raingauge stations Raingauge station
Raingauge value [mm]
Maximum radar value [mm]
Area
Warburg–Welda Bad Driburg–Herste (daily value) Wiehl (daily value) Hennef–Stadt Blankenberg (daily value) Niederdielfen KLG Beckum KA (daily value) Soest KLG
74.5 81.8 110.5 73.5 115.4 86.0 35.1
100 110 138 104 134 96 64
Warburg Bad Driburg Gummersbach Eitorf Siegen Ahlen–Beckum Soest-Welver
346
M. Jessen et al. / Atmospheric Research 77 (2005) 337–346
Further results of this work are: ! preparation of catchment and pixel time series (time series for each radar pixel of the catchments), ! classification of the events on the basis of extreme value statistics (DWD, 1997), ! maps with the damage locations for each event, ! outlets, bridges and other buildings near rivers and creeks are the problematic points for damages.
6. Consequences Consequences for the use of radar data are: ! the quality of the radar images varies between the different radar stations and has to be checked and–where feasible–corrected carefully, ! for the adjustment of radar data, a high density of raingauge stations (continuous and daily values) is helpful. The produced catchment and pixel time series will be used in hydrological models for planning purposes and for scenario simulations of extreme rainfall.
References DWD, 1997. AKORD—Anwenderorientierte Organisation von Radardaten. Deutscher Wetterdienst, Offenbach. Einfalt, T., Arnbjerg-Nielsen, K., Faure, D., Jensen, N.E., Quirmbach, M., Vaes, G., Vieux, B., Golz, C., 2004. Towards a roadmap for use of radar rainfall data in urban drainage. Journal of Hydrology 299, 3 – 4. Einfalt, T., Golz, C., Jessen, M., 2004. Searching for rainfall truth: multisensor thunderstorm analysis. 3rd ERAD, Visby, Sweden, 6–10 September. Harrison, D.L., Driscoll, S.J., Kitchen, M., 2000. Improving precipitation estimates from weather radar using quality control and correction techniques. Meteorological Applications 6, 135 – 144. Joss, J., Germann, U., 2004. Hourly Assessment Factors between Radar and Raingauge, II. VOLTAIRE workshop, Ljubljana, October. Meischner, P., 2003. Weather Radar: Principles and Advanced Applications. Springer Verlag. Sempere-Torres, D., Sa´nchez-Diezma, R., Cordoba, M.A., Pascual, R., Zawadski, I., 2001. An operational methodology to control radar measurements stability from mountain returns. 30th Conf. On Radar Met., Munich, Germany, pp. 264 – 266. Wilson, J.W., Brandes, E.A., 1979. Radar Measurement of Rainfall—A Summary, vol. 60. American Meteorological Society, pp. 1048 – 1058.